Skip to content
Agent YAP

What is it?

Verifier

A verifier is a model whose job is to check work rather than produce it. In Chapter 3, you generate many candidate solutions to a hard problem and use a verifier to score them and pick the best. A good verifier is hard to fool, which is what makes this approach powerful: even if most of a model's attempts are flawed, a strong verifier can reliably find the good ones.

From the Planning and Reasoning glossary

Where Verifier is explained

  1. Subagents and Multi-Agent OrchestrationBuilding Coding Agents and Harnesses · The orchestrator and verifier patterns
  2. Let's Verify Step by Step, Rewarding Good ReasoningPlanning and Reasoning · How a verifier makes a model smarter
  3. Why Multi-Agent Systems Fail: The MAST TaxonomyMulti-Agent Systems · The three categories of failure
  4. Code Review and Security AgentsBuilding Coding Agents and Harnesses · The lesson: an agent cannot mark its own homework
  5. Observability, Hooks, and the Agent SDKBuilding Coding Agents and Harnesses · Concept explanation

Every term, A to Z